In message <[EMAIL PROTECTED]>,Atro Tossavainen writ
es:
>I have a weird problem with an SGI Indy. If I try to start any programs
>that reside on AFS, the processes hang and remain there despite kill -9's.
is it an indy r5000 rev 1.0?
>program (say, pagsh) after copying it to the local disk works. What
>gives? This does make AFS quite unusable on this machine.
can you still run o32 binaries?
if you answered yes to the above two questions i have seen the same
problem with a couple of the indy's here. there is a fixup in the
kernel for hw bugs in the rev 1.0 silicon that interacts with afs.
the kernel, when mapping in pages, wants to remap the text pages
of n32 executables in order to correct an instruction bug in the rev
1.0 r5000. while doing this is builds a hash of the mapped text pages
using fid2() or fid(). afs returns EINVAL for fid2() which prevents the
kernel from trying fid() apparently this is to prevent people from
using the checkpoint facility, which also uses fid2(), with afs files.
however, this also keeps the kernel from being able to map in the pages.
the workaround? we have afs src. i didnt file a pr since i didnt
see anyone else experiencing this problem. besides, i usually
dont get a response when i fix the problem myself.
